ABSTRACT
Objective To establish mathematical model for the differential diagnosis of endoscopic gastric ulcerative lesions with the help of image processing software and statistical analysis method. Methods The microscopic images of 25 gastric ulcers and 25 gastric peptic cancers were collected from January 2013 to December 2013.Then the Photoshop CS4 software was used for image processing and representative measurements gathering(area,perimeter,roundness,height,width,gray scale,involving the density,color,lightness,pixels).Then the principal component analysis,the scatter plot and cluster analysis were performed by SPSS 16.0.Results According to the results of principal component analysis, Z1 showed a strong positive correlation with area,perimeter,height,width,cumulative density,pixel,which represented the shape statistics of ulcer.And Z2 showed a strong positive correlation with gray scale,color, lightness,which represented the color statistics of ulcer.On the Scatter plot,gastric peptic cancer images showed a stronger positive correlation with the principal component Z1 and Z2 .Clustering analysis results showed that for the majority of patients,there was a pattern in its incidence.In the gastric ulcer diagnostic models,the effective information should include area≤10 000;perimeter≤250;height≤90;width≤60;gray scale≥90;cumulative density≤190 000;color≥110;and lightness≥110(unit:pixel).In the gastric ulcera-tive cancers,the effective information should include area≥38 000,perimeter≥690,height≥300,width ≥ 426,gray scale ≤ 100,cumulative density ≥ 14 000 000,color ≤ 125,and lightness ≤ 130 (unit:pixel). Conclusion Among the endoscopic image data,area,perimeter,roundness,height,width,gray scale, cumulative density,color,brightness,pixels are important indices,which can contribute to the differential diagnosis of gastric ulcerative lesions.The established mathematical model based on the ten indices can be applied to predict and diagnose gastric ulcerative lesions.
ABSTRACT
Objective To establish a new approach based on mathematical morphology that can effectively reduce the drifts in functional magnetic resonance imaging (fMRI) signals. Methods Based on investigation of the characteristic of drifted fMRI signals, a mathematical morphology method for baseline drift correction was presented. Results With both simulated data and real fMRI data, the results of the experiment showed that the mathematical morphology method can effectively correct the baseline drifts. Conclusion Both linear and nonlinear drifts can be removed with the proposed method without any statistical model assumption.
ABSTRACT
AIM: To find a more efficient and accurate method for medical image enhancement to resolve the problem of low local contrast, which often appears in magnetic resonance (MR) images.METHODS: Based on the investigation into the previous methods, a new algorithm was presented. The main features were as follows: Non-dual was used in the morphological operations and addition instead of multiplication was used in the contrast stretching operations. To avoid some gray-level bias, the method of normalized gray-level under condition was proposed.RESULTS: The new algorithm was tested by real MR images and simulated experiments. Compared with the previous method, the new method is more accurate, faster and less sensitive to noise.CONCLUSION: Non-dual morphological operations can achieve local contrast enhancement in a more accurate way, and the gray level bias can be eliminated by normalization.